CCRL Discussion Board

https://kirill-kryukov.com/chess/discus ... f=6&t=7258
Perhaps you should remind Dann Corbit

Looks like you can send him an email through this forum.

They probably exist, but I don't remember seeing one. It is clear that KRPPvKRP is the most important table. To play out KRPPvKRP endings reliably, you need all tables that can be reached from KRPPvKRP by one or more promotions (both .rtbw and .rtbz). But if you just want to probe win/draw/loss, it ...

There are *.rtbw and *.rtbz files. The *.rtbw files contain WDL (win/draw/loss) information. The *.rtbz files contain DTZ (distance to zero) information. Distance to zero means the minimum (if you're winning) or maximum (if you're losing) number of moves until a mate, capture or pawn move. The *.rtb...

syjytg wrote:What does broken positions mean in Nalimov Tablebase?

Values of the index range that correspond to illegal positions.

One thing I have been putting off, but which I can no longer avoid is e.p. captures. :) I have been thinking about just ignoring them during generation, and then do a two (or three) position lookup when an e.p. square is present in the lookup request. Is that the way to handle it, or is there somet...

One more thing, though: During generation I am also encoding multiple pieces of the same type and color (up to 4 pieces) as one "piece configuration", that shouldnt matter either, as far as I can tell, or am I wrong here ? For, say, KRRvKQ it is certainly possible to use an indexing schem...

I am using the a1-d1-d4-a1 triangle scheme when doing endgames without pawns. I am also employing that scheme during generation to save space. I am wondering if that is a bad idea. No, that is a good idea. I was having problems with the reflection on the diagonal, so now I, in addition to updating ...

Completely unrelated to that problem, but also on the subject of equal material on both sides: If you have the KQ vs. KQ class generated, isnt it possible to rely on the wtm table exclusively, then if you have to probe a btm position, simply flip the board+ep. square , swap the colors for all piece...

I think your second method is in principle correct. Your algorithm can still be improved though. One improvement is to not immediately check whether a position that might be lost is actually lost, but mark such positions as "potential loss" and check them for being a real loss in the next ...

Note that if a version of the Troitsky line rule reading: If the black pawn stands on or behind the Troitsky line and a white knight stands immediately in front of the pawn and is protected by the other knight, then the position is won for White (and the reverse in colours) were implemented this wo...

I would guess from your previous reply that you know of no analysis of the "second Troitzky line" result (with a specific meaning of "securely blockaded" such as the one I suggest) using a DTM50 database. From the very limited analysis I have done myself (with the aid of the Nal...

I agree that "securely blockaded" is unclear. I think the only good definition is "a knight blockades (stands directly in front of) the pawn and is guarded by the other knight ". But would that not greatly diminish the significance of the second Troitzky line? For the first Troi...

I'm guessing that "securely blockaded" means that the knight cannot easily be chased away by the enemy king. Unfortunately that is not a very clear definition.

How about this answer: When computers have 64 times as much ram, i.e. 64*16 GB = 1 TB of ram. So, perhaps in 12 years assuming doubling every 2 years? If we have a doubling every 2 years, then 5 more years should suffice. I bought my current PC with 64 GB almost 3 years ago. But I don't expect to b...

My guess is that the sign is from the point of white. So white loses, black wins. Therefore -3 is listed on top: only 3 moves to a winning conversion.

What about the endgame tablebase generator reordering the drawing moves during generation of the files? That's not how tablebases work. They don't store moves. They only store values, e.g. mate-in-10 or draw. Given a chess position, the probing code calculates an index number N and then looks up th...

Engines are not running during tablebase look-up anyway, so what would be the loss really? My engine runs (and so does SF), as it does what I propose above :) What I'm not sure of is who would write that code, the Chessbase programmers, or the endgame tablebase programmers. My suggestion would be i...

I wasn't proposing to store recommended drawing moves in a database. Much prefer the forward-search with backed-up 'probability of being chosen' instead. That's how I read the OP's suggestion. Or at least that's how I would try to implement counting the number of paths. Calculating recursively with...

Often there is only one move out of many possible moves that will maintain the draw. And often that is the obvious move that any newbie would play anyway. I don't think trying to derive information from the number of drawing paths is going to work well. In addition, even if it is possible to comput...

I just came across this thread which seems to be along similar (but more general) lines:

http://mathoverflow.net/questions/63423 ... ega-moves/

Can someone please explain why Stockfish 6x64 Modern is reporting tablebase hits with more than 5 peices on the board? At one point there where 12 peices in play and it reported tens of thousands of hits, how can this be possible with 3-4-5 man TBS in operation, the bases seem to be operating ok as...

Sorry for my previous reply, but maybe you can imagine the number of times this question has been asked before and, often, answered. (And I really do not want to play the recurring "are they there yet?" game, even if it's only once a year. I work on the things that I choose to work on.)

JohnnyBob wrote:When?

To avoid recurring themes: never. So now you can stop asking (and everybody else, don't bother, thanks).

Can you estimate how much RAM and time is needed for 7-men tables generated? About 1 TB of RAM. Generation time probably about 64x as long per table. The current generator would need a few modifications to be able to generate 7-men tables. Mostly some internal limits, which is easy, but something n...

from the practical point of wiew, is it justified for black to memorize the winning tactics for the Philidor, for all displacements of this type? I don't know how much of a beginner you are, but from a practical point of view I am pretty sure spending time on QvR is not justified. Unless you're not...